Plotly Express

Check the presentation blog post Introducing Plotly Express.

# Plotly Express is now a sublibrary of Plotly:
import plotly.express as px

Tidy DataFrames as input in Plotly Express

Tidy Data according to Hadley Wickham: (original paper)

• Each variable forms a column and contains values
• Each observation forms a row
• Each type of observational unit forms a table

import pandas as pd

# Here is a "messy" dataframe (wide-form):
messy = pd.DataFrame({
    'patient': ['John Smith', 'Jane Doe', 'Mary Johnson'],
    'Treatment A': [None, 16, 3],
    'Treatment B': [2, 11, 1],
})

messy

	patient	Treatment A	Treatment B
0	John Smith	NaN	2
1	Jane Doe	16.0	11
2	Mary Johnson	3.0	1

# The pandas method `.melt` can often be used to tidy the data (long-form):
tidy = messy.melt(
    id_vars=['patient'],
    value_vars=['Treatment A', 'Treatment B'],
    var_name='Treatment',
    value_name='Result',
)

tidy

	patient	Treatment	Result
0	John Smith	Treatment A	NaN
1	Jane Doe	Treatment A	16.0
2	Mary Johnson	Treatment A	3.0
3	John Smith	Treatment B	2.0
4	Jane Doe	Treatment B	11.0
5	Mary Johnson	Treatment B	1.0

# Once it is in tidy format, plotly express allows you to build complex interactive plots with a one-liner:
px.bar(
    data_frame=tidy,
    x='patient',
    y='Result',
    color='Treatment',
    barmode='group',
    title='Medical Treatment Results',
)

IMPORTANT: From Plotly.py version 4.8 it is no longer a requirement to use tidy dataframes. See this post.

# Plotly express also takes messy dataframes (suitable for quick visualizations during data exploration).
px.bar(
    messy,
    x='patient',
    y=['Treatment A', 'Treatment B'],
    barmode='group',
    title='Medical Treatment Results',
)

Exercise 1

Use the example cluster data, loaded with:

import pandas as pd
table = pd.read_csv(
    'https://raw.githubusercontent.com/chumo/Data2Serve/master/transition_clusters.csv')

... and convert it into a tidy dataframe (HINT: use pd.concat method). It should look like this:

table.head()

tidy = pd.concat([
    table
    .loc[:, ['Xi', 'Yi', 'color']]
    .assign(initial=True)
    .rename(columns={'Xi': 'x', 'Yi': 'y'}),

    table
    .loc[:, ['Xf', 'Yf', 'color']]
    .assign(initial=False)
    .rename(columns={'Xf': 'x', 'Yf': 'y'})],
)

tidy

	x	y	color	initial
0	109.360643	434.557514	red	True
1	55.957358	438.934136	red	True
2	369.115969	419.904538	red	True
3	491.392739	492.316412	red	True
4	34.286602	404.017801	red	True
...	...	...	...	...
85	298.802341	149.868698	blue	False
86	322.576822	108.178118	blue	False
87	293.644438	130.646623	blue	False
88	304.092241	91.984033	blue	False
89	320.828650	119.716195	blue	False

180 rows × 4 columns

Exercise 2

Take the tidy dataframe created in Exercise 1 and build this plot with two subplots:

import plotly.express as px

px.scatter(
    data_frame=tidy,
    x='x',
    y='y',
    color_discrete_map={'red':'red', 'green':'green', 'blue':'blue'},
    color='color',
    facet_row='initial',
    labels={'initial': 'better'},
    width=600,
    height=1000,
    title='Some random data',
).update_layout(
    showlegend=False
)

And now making each subplot as a frame of an animated plot:

import plotly.express as px

px.scatter(
    data_frame=tidy.assign(initial=tidy.initial.astype(str)),
    x='x',
    y='y',
    color_discrete_map={'red':'red', 'green':'green', 'blue':'blue'},
    color='color',
   # facet_row='initial',
    labels={'initial': 'better'},
    width=600,
    height=600,
    title='Some random data',
    animation_frame='initial',
).update_layout(
    showlegend=False
).show()

Exercise 3

Using the gapminder data:

gapminder = px.data.gapminder()

gapminder.head()

	country	continent	year	lifeExp	pop	gdpPercap	iso_alpha	iso_num
0	Afghanistan	Asia	1952	28.801	8425333	779.445314	AFG	4
1	Afghanistan	Asia	1957	30.332	9240934	820.853030	AFG	4
2	Afghanistan	Asia	1962	31.997	10267083	853.100710	AFG	4
3	Afghanistan	Asia	1967	34.020	11537966	836.197138	AFG	4
4	Afghanistan	Asia	1972	36.088	13079460	739.981106	AFG	4

Build the following plots:

and this animated plot:

gapminder.columns
gapminder['pop'].describe()

count    1.704000e+03
mean     2.960121e+07
std      1.061579e+08
min      6.001100e+04
25%      2.793664e+06
50%      7.023596e+06
75%      1.958522e+07
max      1.318683e+09
Name: pop, dtype: float64

import plotly.express as px
import numpy as np

px.scatter(
    data_frame=gapminder.loc[gapminder.loc[:,'year'] == 1977,:],
    x='gdpPercap',
    log_x=True,
    y='lifeExp',
    size='pop',
    size_max=70,
    color='continent',
)

import plotly.express as px
import numpy as np

px.box(
    data_frame=gapminder.loc[(gapminder.loc[:,'year'] == 1977) | (gapminder.loc[:,'year'] == 1997),:],
    x='continent',
    y='gdpPercap',
    facet_col = 'year',
    points=False,
    width=1000,
    height=500,
)

import plotly.express as px

px.bar(
    data_frame=gapminder.loc[gapminder.loc[:,'continent']=='Europe',:].sort_values('gdpPercap'),
    x='gdpPercap',
    y='country',
    range_x=[0,gapminder.query('continent == "Europe"').gdpPercap.max()],
    category_orders={'year':list(range(1950,2010))},
    width=600,
    height=600,
    animation_frame='year',
    animation_group='country'
)

Plotly Express powered backend for Pandas plotting

Since Pandas 0.25 it is possible to provide any backend for the .plot plotting API.

Plotly backend (see here) can be set with:

import pandas as pd
pd.options.plotting.backend = 'plotly'

pd.__version__

'1.3.4'

Then you can use the most common functionalities of Plotly Express by passing your Plotly parameters to the .plot method available to any pandas DataFrame;

tidy.plot.bar(
    x='x',
    y='y',
    color='color',
    barmode='group',
    title='Random Stuff',
)